Relevant Term Suggestion Based on Automatic Learning: Application to Cardiovascular Disease Information Retrieval

當一般民眾藉由搜尋引擎在網際網路上搜尋醫療健康資訊時，礙於對醫學詞彙的認知有限，無法使用精確的查詢來描述需求，導致搜尋的結果並不理想。故本研究使用查詢詞推薦來輔助使用者建立更精確的查詢，讓使用者能找到需要的資訊。我們藉由自動學習的方法，自心血管疾病語料庫訓練出相關詞。當使用者輸入一個查詢詞後，系統會列出與此查詢詞相關的詞，讓使用者選擇有用的詞來擴展查詢。我們評估系統推薦的查詢詞與原始查詢詞的相關程度，實驗結果發現系統建議的前10個詞中，約有76.7%是與原查詢詞很相關的。此結果顯示本系統可以提供豐富的相關查詢詞供使用者選擇，以建立更好的查詢。
Internet has become an important medium for health consumers to search for medical health information. Due to the limited knowledge of medical vocabulary, consumers usually use short and general queries which cannot reflect their information needs. To help consumers search for cardiovascular disease information, we adopt query term suggestion to assist consumers in constructing more specific queries. The suggested terms are automatically learned from a corpus consisting of documents about cardiovascular diseases. Experimental results showed that in the top 10 recommended terms, about 76.7% are relevant to the original query term. It shows that our system can provide rich relevant terms to help users construct more accurate queries.